On Mon, 05 Apr 2004 11:55:17 -0500, Wes Palmer
<[EMAIL PROTECTED]> wrote:
>> match, there's no need to fetch the heap tuple. Unfortunately the
>> planner doesn't believe that this is possible:
>> /* Don't believe estimates less than 1... */
>> if (tuples_fetched < 1.0)
>> tuples_fetched = 1.0;
>
>
> Is there an index on message_recipients(Message_Key)?
Yes.
> How many rows satisfy
>
> SELECT * FROM messages WHERE Message_Date BETWEEN ... AND ... ?
db=> select count(*) from messages where message_date between '29-Mar-04'
and '31-Mar-04 23:59:59';
count
737873
(1 row)
Other c
On Fri, Apr 02, 2004 at 11:08:21AM -0600, [EMAIL PROTECTED] wrote:
> db=> explain analyze select count(*) from messages m, message_recipients r,
> addresses a WHERE r.Message_Key=m.Message_Key AND
> a.Address='[EMAIL PROTECTED]' AND a.Address_Key=r.Recipient AND (
> (m.Message_Date >= '29-MAR-04')
On 4/1/04 6:00 PM, "Manfred Koizar" <[EMAIL PROTECTED]> wrote:
> And what are your settings for random_page_cost, effective_cache_size,
> and sort_mem?
I didn't read close enough last time. Random_page_cosst, cpu_tuple_cost,
cpu_index_tuple_cost, and cpu_operator_cosst are all at default.
Effect
